RFC-0003: Multiverse analyses — in-file universes and cross-universe artifact references - #65
Open
EiffL wants to merge 12 commits into
Open
RFC-0003: Multiverse analyses — in-file universes and cross-universe artifact references#65EiffL wants to merge 12 commits into
EiffL wants to merge 12 commits into
Conversation
…se references Drafts the RFC for the change discussed in #52, grounded in the astra-multiverse-example prototype (Steegen et al. 2016 Figure 1 reproduction). Also syncs the RFC index status for RFC-0002 (Accepted). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
… diffs Universes become diffs against an implicit all-defaults baseline (new 'from:' slot, partial decisions); multiverses gain a base/vary generator form alongside enumeration and '*'; formal reference grammar and field tables added. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
Universes are declared in full or as explicit diffs of one another (from: required for partial declarations); generators require an explicit base when vary is partial; Decision.default is demoted to a presentational hint. Records the rejected implicit-baseline design and grounds the principle in the multiverse/PCS stance (Yu & Kumbier 2020). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
- @ qualifier restricted to Output references (external Inputs don't vary)
- universe scoping defined recursively: decisions or unqualified refs to
scoped artifacts propagate it; @ cuts it; invariant recipes forbid
{decisions.*}
- multiverses are sets; when-deactivated decisions make the space a tree
(inactive decisions omitted from selections and derived ids)
- shared universe/multiverse id namespace; declared ids may not spell a
different selection's derived id
- only in-file universes are addressable; enumeration accepts derived ids
- validator checklist synced with all of the above
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
Contributor
|
'from' and 'base' were one concept with two names; worse, ASTRA's existing 'from' means pure re-export aliasing (no overrides), while universe diffing is inherit-and-override. Rename Universe.from to base, note the deliberate contrast with the alias idiom, and state the generator equivalence: vary is decisions generalized to option sets. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
… ids The option-join id grows unboundedly for '*' over wide decision spaces and leaks into every artifact path; making it addressable also forced collision/shadowing rules. Identity now rests solely on the effective selection; addressing a grid point means declaring it (a three-line base diff); storage spelling (join vs digest) becomes a runner convention. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
…ences Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
Adopt '#' as the selector sigil (reserving '@' for a future sources RFC), regex multiverse membership, nested multiverses, union of selection and generation, scalar option selectors, and the tree-wide shared namespace. Remove the universes/ directory entirely (breaking; major bump). Keep the no-privileged-universe stance (explicit base, no defaults completion; 'baseline' is a naming convention only) and the '.' hierarchy separator, recording both divergences in Questions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
…into Questions Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT
EiffL
marked this pull request as ready for review
August 5, 2026 14:25
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds RFC-0003: Multiverse analyses — in-file universes and cross-universe artifact references (
status: Draft), Step 2 of the RFC process for the change discussed in #52. Reconciled with the alternative draft posted on the tracking issue.Summary
ASTRA can describe a multiverse but cannot analyze one: universes live outside
astra.yaml, and everyOutput.inputsedge implicitly means "same universe as me." This RFC proposes:universesinastra.yaml— universes become addressable elements, declared in full or as explicit diffs of one another (base:+ partialdecisions). Theuniverses/directory is removed (breaking; migration is a mechanical move).multiverses— named sets of universes, via selection (universes:— exact ids, nested multiverses, regex over declared ids), generation (base+decisionsoption selectors;decisions: "*"for the full valid space), or the union of both.artifact#selectorreferences — pin an input to one universe (f1_score#svm_standard) or fan out across a set (religiosity_p#full_multiverse), with universe scoping inferred recursively and cut by the#selector.@is reserved for a future sources RFC (revisions), so the axes compose.Design principles: no universe is privileged by the spec — no implicit defaults-derived baseline, no reserved ids (naming a universe
baselineis author convention); generated universes are anonymous — identity is the effective selection, and addressing a grid point means declaring it. The.hierarchy separator is kept.Grounded in the working prototypes
astra-multiverse-exampleandastra-multiverse-example-steegen-2(@anthonyozerov).Main open questions (see the RFC's Questions or objections)
decisions: "*".base) and the:separator (rejected here).Discussion happens here while the PR is a draft, per the RFC process.
Tracking issue: #52
🤖 Generated with Claude Code
https://claude.ai/code/session_01Ah3BGphV5NgkGEhkR3yZDT